Skip to content
Docs
Manage Experiments

Split URL tests

A split URL test sends visitors to different addresses instead of changing the page they are on.

On this page 6

A split URL test sends visitors to different addresses. Some see the page they asked for, the rest are redirected to a page you built separately.

Use it when the new version is its own page. An A/B test edits the page a visitor is already on, which is quicker and does not move anybody. Once the change is a rebuilt page, a different template or a different app, there is nothing for an editor to edit, and this is the test you want.

You need both pages live and reachable before you start. This test does not build the second page for you.

The six steps

Only the first two differ from an A/B test. The rest are the same screens in the same order.

Step Name What you do there
1 Setup & original page Name it, and give the address visitors arrive at.
2 Traffic & destinations List the addresses, and split the traffic.
3 Goals & metrics Same as an A/B test.
4 Audience & targeting Same as an A/B test.
5 Configuration Same as an A/B test.
6 Preview & confirm Same as an A/B test.

For steps three to six, see Your first A/B test.

How to set one up

  1. Setup & original page. Type a name, then the Original URL. That is the address visitors arrive at, and the one some of them are sent away from.
  2. Decide whether to Carry query parameters and Carry the anchor across the redirect.
  3. Traffic & destinations. Every row is an address, the original included. Add a row for each new page, give it a Destination name and its address, and set the splits.
  4. Pick the baseline. It is the original unless you move it.
  5. Work through steps three to six as you would for an A/B test, then press Start.

There is no separate page targeting to write. The test runs wherever the original address is.

What each label means

These are the labels this test has and an A/B test does not. For the others, see Your first A/B test.

Step Label What it means
1 Original URL The address visitors arrive at. A full address including https://. Up to 2048 characters.
1 Carry query parameters Whether ?utm_source=newsletter follows the visitor to the new page.
1 Carry the anchor Whether #compare-plans follows them.
2 Destination name Shown on every report, in place of a variation name. Up to 80 characters.
2 Address Where that share of the traffic is sent. A full address including https://. One row per version.

Turn on Carry query parameters unless you have a reason not to. A campaign that paid for the visit stays attributed. Where the original and the destination both set the same parameter, the destination's value wins.

Leave Carry the anchor off unless you know you want it. An anchor names a section of the original page, and your new page may not have that section.

A test needs at least two rows and takes at most ten. The share of matching traffic goes down to 0.1%.

Before it can start

The checklist at step six is shorter than an A/B test's. A split URL test has no authored changes, so the four checks about changes resolving on the page do not apply. What is left is the snippet, the deciding goal, whether your goals are receiving data, and traffic sharing if the test is in a group.

Good to know

  • The tag has to be on every page in the test, not only the original. A visitor sent to a page with no tag is a visitor whose purchase nothing records. See Install the tag.
  • If a destination is on another host, add that host too. Subdomains count as separate addresses. See Add your domain.
  • The page is held back for a moment while the redirect is worked out, so the visitor does not see the original flash past first. That is why the tag belongs as high in the page as it will go.
  • Each visitor is sent to the same address every time. Somebody who comes back on the same browser does not get shuffled to a different page.
  • Search engines see two pages with much the same content. Point the canonical tag on each new page back at the original, or you are competing with yourself.
  • The address is what the report is about, so name your destinations for what changed, not "Variation B". "Pricing with annual first" reads in six weeks.